How to structure prompt instructions

Numbered steps, delimiters, sections and logical order: how to organize a long prompt so the AI does not skip anything.

A long, messy prompt is like a brief given in a hurry over the phone: something gets lost. Structure makes instructions easy to follow, for the AI and for whoever reuses the prompt.

Use sections

Split the prompt into blocks with clear headings:

## Role
## Context
## Task
## Steps
## Constraints
## Output format

You can use Markdown headings, capitals or tags; what matters is consistency.

Number the steps

If the task has a sequence, number it. Models follow “1, 2, 3” better than a paragraph with “first… then… and also…”.

Delimit the data

Any text the AI must process (not obey) should be delimited:

Translate the text between the tags into French.
<text>
...
</text>

This also reduces the risk that instructions hidden in a pasted document change the AI's behavior.

Positive instructions

“Write in short sentences” works better than “do not write long sentences”. When something is forbidden, say so explicitly, but pair it with the alternative.

Explicit output format

Describe exactly how you want the answer:

Reply with:
1. A 2-line summary.
2. A table with columns: Option | Pros | Cons | Cost.
3. Your recommendation in one paragraph.

For technical integrations you can ask for JSON with specific keys and an example.

Check for contradictions

Before sending, reread: are there conflicting instructions? Is there anything the AI has to guess? One minute of review saves several iterations.

Learn prompting

  • What is a prompt and how does it work

    A prompt is the instruction you give an artificial intelligence. Learn its parts, why the result depends so much on how you write it, and how to get started.

  • How to create a prompt from scratch

    A 6-step method to go from a vague idea to a clear, reusable prompt, with a complete end-to-end example.

  • How to write effective prompts

    Ten practical principles to write prompts that work on the first try more often: clarity, specificity, examples, format and review.

  • How to write prompts for ChatGPT

    Practical tips to get more out of ChatGPT: message structure, custom instructions, long conversations, files and reviewing answers.

  • How to write prompts for Claude

    How to write prompts for Anthropic's Claude: direct instructions, rich context, long documents, tags for structure and requests for reasoning.

  • How to write prompts to generate images

    The structure that works best in Midjourney, DALL-E, Stable Diffusion, Flux or Firefly: subject, style, light, composition and parameters, with examples.

  • How to write prompts to generate videos

    How to describe a shot for Runway, Kling, Sora or Veo: one subject, one action, camera movement, light and mood. And how to plan multi-shot videos.

  • How to improve a bad prompt

    A diagnostic process to find out why a prompt is not working and how to fix it, with a before-and-after example and a prompt to let the AI help you.

  • What are variables in a prompt

    Variables turn a prompt into a reusable template. Learn how to define them, name them and use them on HomeCasa Prompts.

  • What is context in a prompt and how much to give

    Context is the information the AI cannot guess. Learn what to include, what to avoid and how to organize documents and data inside a prompt.

  • How to define roles in a prompt

    What telling the AI “act as…” really does, how to choose a good role, combine it with concrete criteria and avoid roles that do not help.

  • Good and bad prompt examples

    Eight comparisons of real prompts for text, code, image and video, explaining what fails in the bad version and what the good one improves.

  • Common mistakes when writing prompts

    The ten most common prompt-writing mistakes and how to avoid them: from missing context to blindly trusting the answer.

  • Advanced prompting techniques

    Few-shot examples, step-by-step reasoning, prompt chaining, self-critique, structured outputs and evaluation: techniques for more reliable results.